Is the reduced primary care use of antibiotics resulting in reduced antimicrobial resistance: a population-based ecological study

Problem

Antibiotic resistance is considered one of the greatest threats to public health. Primary care is responsible for around 75% of antibiotics prescribed and was therefore the target for an NHS England quality premium to reduce overall antibiotic prescribing, and the percentage of broad-spectrum antibiotics (co-amoxiclav, cephalosporins and quinolones) to under 10%. We aimed to assess if practice level antibiotic prescribing reductions have resulted in practice level changes in resistance patterns in bacteria isolated from routinely submitted urine samples.

Approach

We collected antibiotic resistance data from E. coli urinary tract infection-confirmed urine samples submitted by >140,000 patients registered at 174 practices served by Southmead Hospital and Bristol Royal Infirmary between 2013 and 2016. Data on the total number of antibiotic items dispensed for each practice were extracted from NHS Digital. Multilevel modelling was used for within quarter dispensed drug-bug pairs, taking account of potential confounders including age, sex, socioeconomic status and urban/rural classification. As there could be delayed effects, a second set of analyses investigated resistance in the next quarter.

Findings

More than 85% of patients were female, and over 60% were aged 50 years or older. Reductions in dispensing were observed for several antibiotics including broad-spectrums, an increase in dispensing was observed for nitrofurantoin between 2013 and 2016. Changes in within-quarter antibiotic resistance were relatively small for all antibiotics. Previous quarter reductions in amoxicillin dispensing were associated with reduced amoxicillin resistance (OR 0.997, 95% CI 0.995-0.999, p-value <0.01); increased nitrofurantoin dispensing was associated with reductions in trimethoprim resistance (OR 0.991, 95% CI 0.986-0.996, p-value <0.001), while reductions in broad-spectrum dispensing in the previous quarter increased subsequent broad-spectrum resistance (OR 1.007, 95% CI 1.003-1.011, p-value <0.001).

Consequences

For many antibiotics, dispensing decreased year-on-year between 2013 to 2016, with resistance rates remaining stable. We found some encouraging evidence to suggest that reducing dispensing of certain antibiotics can reduce resistance, in both the preceding quarter and within the same quarter. However, we also found evidence that broad-spectrum resistance continued to increase despite reductions in dispensing. These relationships did not appear to strengthen or weaken over time. Further research is needed to understand the reasons and to assess if they are being replicated nationally.